CN110851729A - Resource information recommendation method, apparatus, device and computer storage medium - Google Patents
Resource information recommendation method, apparatus, device and computer storage mediumDownload PDFInfo
- Publication number
- CN110851729A CN110851729ACN201911145222.1ACN201911145222ACN110851729ACN 110851729 ACN110851729 ACN 110851729ACN 201911145222 ACN201911145222 ACN 201911145222ACN 110851729 ACN110851729 ACN 110851729A
- Authority
- CN
- China
- Prior art keywords
- resource information
- historical
- recommendation
- current
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Translated fromChinese
Description
Translated fromChinese技术领域technical field
本发明涉及金融科技(Fintech)技术领域，尤其涉及资源信息推荐方法、装置、设备及计算机存储介质。The present invention relates to the technical field of financial technology (Fintech), and in particular, to a resource information recommendation method, apparatus, device and computer storage medium.
背景技术Background technique
随着计算机技术的发展，越来越多的技术应用在金融领域，传统金融业正在逐步向金融科技(Fintech)转变，大数据中的企业贷款技术也不例外，但由于金融行业的安全性、实时性要求，也对技术提出的更高的要求。例如在金融平台对各个企业推荐资源信息时，往往需要对已有的历史数据进行分析，并根据分析结果向各个企业进行推荐。但目前金融平台的推荐方式一般是基于用户的协同推荐算法或者是基于资源信息的协同推荐算法来进行推荐的，但是由于这两种推荐方式中的推荐算法并没有考虑企业经营状况和资源信息的变化程度等，往往导致推荐给企业的资源信息不准确。因此，如何提高金融平台中推荐算法推荐的准确性成为了目前亟待解决的技术问题。With the development of computer technology, more and more technologies are applied in the financial field, and the traditional financial industry is gradually transforming into financial technology (Fintech), and the corporate loan technology in big data is no exception. Real-time requirements, but also higher requirements for technology. For example, when a financial platform recommends resource information to various enterprises, it is often necessary to analyze the existing historical data, and make recommendations to various enterprises according to the analysis results. However, the current recommendation methods of financial platforms are generally based on the collaborative recommendation algorithm of users or the collaborative recommendation algorithm based on resource information. The degree of change, etc., often leads to inaccurate resource information recommended to enterprises. Therefore, how to improve the accuracy of recommendation algorithm recommendations in financial platforms has become an urgent technical problem to be solved.
发明内容SUMMARY OF THE INVENTION
本发明的主要目的在于提出一种资源信息推荐方法、装置、设备及计算机存储介质，旨在提高金融平台中推荐算法推荐的准确性。The main purpose of the present invention is to propose a resource information recommendation method, device, equipment and computer storage medium, aiming at improving the accuracy of recommendation algorithm recommendation in a financial platform.
为实现上述目的，本发明提供一种资源信息推荐方法，所述资源信息推荐方法包括如下步骤：In order to achieve the above object, the present invention provides a method for recommending resource information, the method for recommending resource information includes the following steps:
获取各资源信息的历史数据，并基于特征选择在各所述历史数据中获取与用户账号具有相关性的多个历史特征值；Obtaining historical data of each resource information, and selecting a plurality of historical feature values related to the user account in each of the historical data based on feature selection;
计算用户账号的当前特征值与各所述历史特征值之间的相似度值，并在各所述相似度值中获取预设数量的目标相似度值；Calculate the similarity value between the current feature value of the user account and each of the historical feature values, and obtain a preset number of target similarity values from each of the similarity values;
基于各所述资源信息确定各所述目标相似度值对应的目标资源信息，并获取各所述目标资源信息中的参数信息；Determine target resource information corresponding to each of the target similarity values based on each of the resource information, and obtain parameter information in each of the target resource information;
基于预设推荐算法对各所述目标相似度值和各所述参数信息进行计算，以获取各所述资源信息对应的推荐指数，并在各所述推荐指数中确定数值最高的目标推荐指数，对所述目标推荐指数对应的资源信息进行推荐。Based on a preset recommendation algorithm, each of the target similarity values and each of the parameter information is calculated to obtain a recommendation index corresponding to each of the resource information, and a target recommendation index with the highest value is determined among the respective recommendation indices, The resource information corresponding to the target recommendation index is recommended.
可选地，所述基于预设推荐算法对各所述目标相似度值和各所述参数信息进行计算，以获取各所述资源信息对应的推荐指数的步骤，包括：Optionally, the step of calculating each of the target similarity values and each of the parameter information based on a preset recommendation algorithm to obtain a recommendation index corresponding to each of the resource information includes:
依次遍历各所述目标相似度值，并确定当前遍历的当前目标相似度值对应的参数信息；Traversing each of the target similarity values in turn, and determining the parameter information corresponding to the current target similarity value currently traversed;
根据预设推荐算法对所述当前目标相似度值和所述当前目标相似度值对应的参数信息进行计算，以获取所述当前目标相似度值对应的资源信息的推荐指数，直至各所述目标相似度值遍历完成。Calculate the current target similarity value and the parameter information corresponding to the current target similarity value according to the preset recommendation algorithm, so as to obtain the recommendation index of the resource information corresponding to the current target similarity value, until each target The similarity value traversal is completed.
可选地，所述计算用户账号的当前特征值与各所述历史特征值之间的相似度值的步骤，包括：Optionally, the step of calculating the similarity value between the current feature value of the user account and each of the historical feature values includes:
获取所述当前特征值对应的当前标准化特征值和各所述历史特征值对应的历史标准化特征值；Obtain the current standardized feature value corresponding to the current feature value and the historical standardized feature value corresponding to each of the historical feature values;
根据所述当前标准化特征值和各所述历史标准化特征值确定所有相似度值。All similarity values are determined according to the current normalized feature value and each of the historical normalized feature values.
可选地，所述获取所述当前特征值对应的当前标准化特征值和各所述历史特征值对应的历史标准化特征值的步骤，包括：Optionally, the step of obtaining the current standardized feature value corresponding to the current feature value and the historical standardized feature value corresponding to each of the historical feature values includes:
基于预设数据处理方式对所述当前特征值进行标准化，以获取所述当前特征值对应的当前标准化特征值；Standardize the current feature value based on a preset data processing method to obtain a current standardized feature value corresponding to the current feature value;
基于所述预设数据处理方式对各所述历史特征值进行标准化，以获取各所述历史特征值对应的历史标准化特征值。Each of the historical feature values is normalized based on the preset data processing method to obtain a historical standardized feature value corresponding to each of the historical feature values.
可选地，所述根据所述当前标准化特征值和各所述历史标准化特征值确定所有相似度值的步骤，包括：Optionally, the step of determining all similarity values according to the current standardized feature value and each of the historical standardized feature values includes:
依次遍历各所述历史标准化特征值，并计算当前遍历的历史标准化特征值和所述当前标准化特征值之间的欧式距离；Traversing each of the historical standardized eigenvalues in turn, and calculating the Euclidean distance between the currently traversed historical standardized eigenvalues and the current standardized eigenvalues;
基于所述欧式距离确定当前遍历的历史标准化特征值和当前标准化特征值的相似度值，直至各所述历史标准化特征值遍历完成。Based on the Euclidean distance, the currently traversed historical normalized feature value and the similarity value of the current normalized feature value are determined until each historical normalized feature value traversal is completed.
可选地，所述基于特征选择在各所述历史数据中获取与用户账号具有相关性的多个历史特征值的步骤，包括：Optionally, the step of acquiring a plurality of historical feature values related to the user account in each of the historical data based on the feature selection includes:
基于预设校验算法对各所述历史数据进行离散化，以获取各所述历史数据对应的初级特征值；Discretize each of the historical data based on a preset verification algorithm to obtain the primary characteristic value corresponding to each of the historical data;
基于所述预设校验算法对各所述初级特征值进行评分，并基于评分结果获取与用户账号具有相关性的多个历史特征值。Each of the primary feature values is scored based on the preset verification algorithm, and a plurality of historical feature values related to the user account are acquired based on the scoring result.
可选地，所述计算用户账号的当前特征值与各所述历史特征值之间的相似度值的步骤之前，包括：Optionally, before the step of calculating the similarity value between the current feature value of the user account and each of the historical feature values, the step includes:
获取用户账号的当前数据，并对所述当前数据进行离散数据编码，以获取所述当前账号的当前特征值。The current data of the user account is obtained, and discrete data encoding is performed on the current data to obtain the current feature value of the current account.
此外，为实现上述目的，本发明还提供一种资源信息推荐装置，所述资源信息推荐装置包括：In addition, in order to achieve the above object, the present invention also provides a resource information recommendation device, the resource information recommendation device includes:
获取模块，用于获取各资源信息的历史数据，并基于特征选择在各所述历史数据中获取与用户账号具有相关性的多个历史特征值；an acquisition module, configured to acquire historical data of each resource information, and to acquire a plurality of historical feature values relevant to the user account in each of the historical data based on feature selection;
计算模块，用于计算用户账号的当前特征值与各所述历史特征值之间的相似度值，并在各所述相似度值中获取预设数量的目标相似度值；a calculation module, configured to calculate the similarity value between the current feature value of the user account and each of the historical feature values, and obtain a preset number of target similarity values from each of the similarity values;
确定模块，用于基于各所述资源信息确定各所述目标相似度值对应的目标资源信息，并获取各所述目标资源信息中的参数信息；a determining module, configured to determine target resource information corresponding to each of the target similarity values based on each of the resource information, and obtain parameter information in each of the target resource information;
推荐模块，用于基于预设推荐算法对各所述目标相似度值和各所述参数信息进行计算，以获取各所述资源信息对应的推荐指数，并在各所述推荐指数中确定数值最高的目标推荐指数，对所述目标推荐指数对应的资源信息进行推荐。A recommendation module, configured to calculate each of the target similarity values and each of the parameter information based on a preset recommendation algorithm, to obtain a recommendation index corresponding to each of the resource information, and to determine the highest value in each of the recommended indices The target recommendation index is the target recommendation index, and the resource information corresponding to the target recommendation index is recommended.
此外，为实现上述目的，本发明还提供一种资源信息推荐设备，所述资源信息推荐设备包括：存储器、处理器及存储在所述存储器上并可在所述处理器上运行的资源信息推荐程序，所述资源信息推荐程序被所述处理器执行时实现如上所述的资源信息推荐方法的步骤。In addition, in order to achieve the above object, the present invention also provides a resource information recommendation device, the resource information recommendation device includes: a memory, a processor, and a resource information recommendation device that is stored in the memory and can run on the processor. A program, when the resource information recommendation program is executed by the processor, implements the steps of the resource information recommendation method described above.
此外，为实现上述目的，本发明还提供一种计算机存储介质，所述计算机存储介质上存储有资源信息推荐程序，所述资源信息推荐程序被处理器执行时实现如上所述的资源信息推荐方法的步骤。In addition, in order to achieve the above object, the present invention also provides a computer storage medium on which a resource information recommendation program is stored, and when the resource information recommendation program is executed by a processor, the above-mentioned resource information recommendation method is implemented A step of.
本发明通过获取各资源信息的历史数据，并基于特征选择在各所述历史数据中获取与用户账号具有相关性的多个历史特征值；计算用户账号的当前特征值与各所述历史特征值之间的相似度值，并在各所述相似度值中获取预设数量的目标相似度值；基于各所述资源信息确定各所述目标相似度值对应的目标资源信息，并获取各所述目标资源信息中的参数信息；基于预设推荐算法对各所述目标相似度值和各所述参数信息进行计算，以获取各所述资源信息对应的推荐指数，并在各所述推荐指数中确定数值最高的目标推荐指数，对所述目标推荐指数对应的资源信息进行推荐。通过获取各个资源信息的历史数据，并将用户账号的当前特征值和历史数据中的多个历史特征值进行比较，以确定相似度值，也就是在计算相似度值时，综合考虑了资源信息的历史数据，提高了向用户账号推荐资源信息的准确性，并且还需要在这些相似度值中确定目标相似度值，并根据参数信息确定目标推荐指数，将目标推荐指数对应的资源信息进行推荐，使得本次推荐的效果更好，更符合用户的需求，提高了金融平台中推荐算法推荐的准确性。The present invention obtains the historical data of each resource information and selects a plurality of historical feature values related to the user account in each of the historical data based on the feature selection; calculates the current feature value of the user account and each of the historical feature values. and obtain a preset number of target similarity values from each of the similarity values; determine the target resource information corresponding to each of the target similarity values based on each of the resource information, and obtain each of the target similarity values. The parameter information in the target resource information; based on the preset recommendation algorithm, the similarity value of each target and the parameter information are calculated to obtain the recommendation index corresponding to the resource information. The target recommendation index with the highest value is determined in the target recommendation index, and the resource information corresponding to the target recommendation index is recommended. By acquiring the historical data of each resource information, and comparing the current feature value of the user account with multiple historical feature values in the historical data, the similarity value is determined, that is, the resource information is comprehensively considered when calculating the similarity value. It improves the accuracy of recommending resource information to user accounts, and also needs to determine the target similarity value among these similarity values, and determine the target recommendation index according to the parameter information, and recommend the resource information corresponding to the target recommendation index. , so that the effect of this recommendation is better, more in line with the needs of users, and improves the accuracy of the recommendation algorithm in the financial platform.
附图说明Description of drawings
图1是本发明实施例方案涉及的硬件运行环境的设备结构示意图；1 is a schematic diagram of a device structure of a hardware operating environment involved in an embodiment of the present invention;
图2为本发明资源信息推荐方法第一实施例的流程示意图；2 is a schematic flowchart of a first embodiment of a method for recommending resource information according to the present invention;
图3为本发明资源信息推荐装置的装置模块示意图；3 is a schematic diagram of a device module of an apparatus for recommending resource information according to the present invention;
图4为本发明资源信息推荐方法中申请产品的流程示意图；FIG. 4 is a schematic flow chart of applying for a product in the resource information recommendation method of the present invention;
图5为本发明资源信息推荐方法中的流程示意图。FIG. 5 is a schematic flowchart of a method for recommending resource information according to the present invention.
本发明目的的实现、功能特点及优点将结合实施例，参照附图做进一步说明。The realization, functional characteristics and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments.
具体实施方式Detailed ways
应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.
如图1所示，图1是本发明实施例方案涉及的硬件运行环境的设备结构示意图。As shown in FIG. 1 , FIG. 1 is a schematic diagram of a device structure of a hardware operating environment involved in an embodiment of the present invention.
本发明实施例资源信息推荐设备可以是PC机或服务器设备，其上运行有Java虚拟机。The resource information recommendation device in the embodiment of the present invention may be a PC or a server device, on which a Java virtual machine runs.
如图1所示，该资源信息推荐设备可以包括：处理器1001，例如CPU，网络接口1004，用户接口1003，存储器1005，通信总线1002。其中，通信总线1002用于实现这些组件之间的连接通信。用户接口1003可以包括显示屏(Display)、输入单元比如键盘(Keyboard)，可选用户接口1003还可以包括标准的有线接口、无线接口。网络接口1004可选的可以包括标准的有线接口、无线接口(如WI-FI接口)。存储器1005可以是高速RAM存储器，也可以是稳定的存储器(non-volatile memory)，例如磁盘存储器。存储器1005可选的还可以是独立于前述处理器1001的存储装置。As shown in FIG. 1 , the resource information recommendation device may include: a
本领域技术人员可以理解，图1中示出的设备结构并不构成对设备的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件布置。Those skilled in the art can understand that the device structure shown in FIG. 1 does not constitute a limitation on the device, and may include more or less components than the one shown, or combine some components, or arrange different components.
如图1所示，作为一种计算机存储介质的存储器1005中可以包括操作系统、网络通信模块、用户接口模块以及资源信息推荐程序。As shown in FIG. 1 , the
在图1所示的设备中，网络接口1004主要用于连接后台服务器，与后台服务器进行数据通信；用户接口1003主要用于连接客户端(用户端)，与客户端进行数据通信；而处理器1001可以用于调用存储器1005中存储的资源信息推荐程序，并执行下述资源信息推荐方法中的操作。In the device shown in FIG. 1 , the
基于上述硬件结构，提出本发明资源信息推荐方法实施例。Based on the above hardware structure, an embodiment of the resource information recommendation method of the present invention is proposed.
参照图2，图2为本发明资源信息推荐方法第一实施例的流程示意图，所述方法包括：Referring to FIG. 2, FIG. 2 is a schematic flowchart of a first embodiment of a method for recommending resource information according to the present invention. The method includes:
步骤S10，获取各资源信息的历史数据，并基于特征选择在各所述历史数据中获取与用户账号具有相关性的多个历史特征值；Step S10, obtaining historical data of each resource information, and selecting a plurality of historical feature values relevant to the user account in each of the historical data based on feature selection;
在本实施例中用户账号可以为企业在金融平台中登录的账号。目前，金融平台为企业推荐各个资源信息(如金融资源信息)的方式主要有：(1)topN：即按对金融平台上每个资源信息申请成功的企业数从多到少的排名，并按排名进行推荐；(2)基于用户的协同推荐算法：主要原理是推荐和用户相似用户曾经成功申请的资源信息，主要步骤如下，找到与目标企业U相似的企业集合：即通过计算企业U和其它每个企业间的相似度找到最相似的企业集合，常用的相似度算法为Jaccard算法和余弦相似度算法：Jaccard算法：即两个企业申请资源信息的交集除以两个企业申请资源信息的并集。In this embodiment, the user account may be an account that the enterprise logs in to the financial platform. At present, the main ways for financial platforms to recommend various resource information (such as financial resource information) for enterprises are: (1) topN: that is, according to the ranking of the number of enterprises that have successfully applied for each resource information on the financial platform, and according to (2) User-based collaborative recommendation algorithm: the main principle is to recommend resource information that users similar to the user have successfully applied for. The main steps are as follows, to find a set of enterprises similar to the target enterprise U: that is, by calculating the enterprise U and other The similarity between each enterprise finds the most similar enterprise set. The commonly used similarity algorithms are Jaccard algorithm and cosine similarity algorithm: Jaccard algorithm: that is, the intersection of the application resource information of two enterprises divided by the union of the application resource information of the two enterprises set.
用户U和V相似度
将企业U和其它企业相似度进行从大到小排序，选择前L个企业；针对这L个企业L＝{L1,L2,…Ln}申请成功的资源信息集合P＝{P1，P2,…Pn}，资源信息Pi对用户的推荐指数C(Pi,U)为：Sort the similarity between enterprise U and other enterprises in descending order, and select the first L enterprises; for these L enterprises L={L1, L2,...Ln}, the resource information set P={P1, P2,... Pn}, the recommendation index C(Pi, U) of the resource information Pi to the user is:
N(pi)为对申请成功过Pi的企业。对资源信息集合P根据每个Pi的推荐指数C(Pi,U)值从大到小进行排序，选取前K个资源信息，得到企业U的信息推荐列表。N(pi) is for companies that have successfully applied for Pi. The resource information set P is sorted from large to small according to the recommendation index C(Pi, U) value of each Pi, and the top K resource information is selected to obtain the information recommendation list of enterprise U.
(3)基于资源信息的协同推荐算法：主要原理是和用户曾经申请成功的资源信息越相似，推荐指数也越高，主要步骤同基于用户的协同推荐算法相似，具体如下：计算资源信息集合P中两两的相似度，即计算各资源信息之间的相似度Sij，相似度算法采用Jaccard算法；计算资源信息P对于用户U的推荐指数C(P,U),选取最高推荐指数的K个资源信息进行推荐，推荐指数C(P,U)算法如下：(3) Collaborative recommendation algorithm based on resource information: The main principle is that the more similar the resource information the user has successfully applied for, the higher the recommendation index. The main steps are similar to the user-based collaborative recommendation algorithm, as follows: Calculate the resource information set P The similarity between the two is to calculate the similarity Sij between each resource information, and the similarity algorithm adopts the Jaccard algorithm; calculate the recommendation index C(P, U) of the resource information P for the user U, and select the K with the highest recommendation index The resource information is recommended, and the recommendation index C(P,U) algorithm is as follows:
L为和资源信息j最为相似的l个资源信息。N(u)为用户U曾经申请成功的资源信息。L is the l resource information most similar to resource information j. N(u) is the resource information that the user U has successfully applied for.
但是由于大部分企业申请成功的资源信息数较少，特别是小微企业许多之前并没有申请成功过资源信息，因此计算相似度时绝大部分为0，基于用户和基于资源信息的协同推荐算法并不能获得很好的推荐效果。并且由于企业经营状况不断变化，资源信息的准入条件也在发生变化，之前的用户申请成功的资源信息及与其类型的资源信息，在经过一段时间后并不一定能申请成功，目前的算法并没有考虑到企业经营状况和产品的变化程度。而topN的算法推荐了申请成功企业最多的产品，但某个资源信息的申请成功的企业数量相比于企业的总量比例很小，并不适用于绝大部分企业，推荐效果不佳。However, since most enterprises have successfully applied for a small number of resource information, especially many small and micro enterprises have not successfully applied for resource information before, most of them are 0 when calculating the similarity. The collaborative recommendation algorithm based on users and resource information And can not get a good recommendation effect. In addition, due to the continuous changes in the business situation of the enterprise, the access conditions for resource information are also changing. The resource information and the type of resource information that the previous users successfully applied for may not be able to apply successfully after a period of time. The current algorithm does not It does not take into account the degree of change in business conditions and products. The topN algorithm recommends the products that have successfully applied for the most companies, but the number of companies that have successfully applied for a certain resource information is small compared to the total number of companies, so it is not applicable to most companies, and the recommendation effect is not good.
因此，在本实施例中，通过根据平台企业基本信息及成功申请的历史数据，并通过特征选择筛选出独立且相关性较大的特征，即基于特征选择在各个历史数据中获取与用户账号具有相关性的多个历史特征值。Therefore, in this embodiment, independent and highly relevant features are screened out through feature selection according to the basic information of the platform company and the historical data of successful applications, that is, based on feature selection, each historical data is obtained and the user account has a Multiple historical eigenvalues for correlation.
并且由于金融平台会记录企业的基本信息和经营信息，以便银行进行初步筛选，市场化的金融平台会根据企业授权获得工商信息，政府化的金融平台可以通过当地征信平台或大数据局获得相关企业工商、税务数据。当企业申请资源信息时，金融平台会记录下企业申请资源信息时最新的工商、税务数据、申请结果等信息，将这些信息进行衍生特征加工可整理为如下表1。表中同一个企业会申请多个资源信息，同一个资源信息也会有多个企业申请，同一个企业也会申请同一个资源信息多次。And because the financial platform will record the basic information and business information of the enterprise, so that the bank can conduct preliminary screening, the market-oriented financial platform will obtain the industrial and commercial information according to the authorization of the enterprise, and the government-oriented financial platform can obtain relevant information through the local credit reporting platform or the big data bureau. Business, business and tax data. When an enterprise applies for resource information, the financial platform will record the latest industrial and commercial, tax data, application results and other information when the enterprise applies for resource information, and these information can be sorted into the following table 1 by processing derivative features. In the table, the same enterprise will apply for multiple resource information, the same resource information will also be applied by multiple enterprises, and the same enterprise will also apply for the same resource information multiple times.
表1Table 1
其中资源信息一般会对企业所处行业进行判断，由于行业代码公分4层，1380个小类，不易区分，因此只取第1层的代码，既每个企业的行业类型只区分到第1层的大类。申请日期为当前日期的5年内。Among them, resource information generally judges the industry in which the enterprise is located. Since the industry code has 4 layers and 1380 sub-categories, it is difficult to distinguish, so only the code of the first layer is taken, that is, the industry type of each enterprise is only distinguished to the first layer. category. The application date is within 5 years of the current date.
数据特征选择主要选择影响申请结果的主要特征，主要步骤如下：The data feature selection mainly selects the main features that affect the application result. The main steps are as follows:
连续数据离散化：对表1中连续型变量通过卡方检验方法进行离散化，如对上年营业收入字段离散化后为：小于等于3000万为1，大于3000万并且小于等于1亿为2，大于1亿并且小于等于10亿为3，大于10亿为4。特征选择：采用过滤法中的卡方检验法对每个特征进行评分，选取最好的特征；通过特征选择，选取M个对企业成功申请资源信息相关性最高的特征。Continuous data discretization: The continuous variables in Table 1 are discretized by the chi-square test method. For example, after the discretization of the operating income field of the previous year, it is 1 if it is less than or equal to 30 million, and 2 if it is greater than 30 million and less than or equal to 100 million. , greater than 100 million and less than or equal to 1 billion is 3, and greater than 1 billion is 4. Feature selection: use the chi-square test method in the filtering method to score each feature, and select the best feature; through feature selection, select M features that are most relevant to the company's successful application for resource information.
步骤S20，计算用户账号的当前特征值与各所述历史特征值之间的相似度值，并在各所述相似度值中获取预设数量的目标相似度值；Step S20, calculating the similarity value between the current feature value of the user account and each of the historical feature values, and obtaining a preset number of target similarity values from each of the similarity values;
当通过特征筛选，并且只选择申请成功的记录，可得如下表2：After passing the feature screening and only selecting the records with successful application, the following table 2 can be obtained:
表2Table 2
并且在企业在平台上基于用户账号进行登录后，也就是对于登录平台的企业，会根据企业当前最新的选择特征与各申请成功的企业特征进行相似度计算，计算步骤如下：对于用户账号的数据信息进行离散数据编码，即对于像纳税人等级的有序序列{A->B->C->D}转换为有序数列如{1、2、3、4}；对于行业等无序序列，采用one-hot编码，即假设有3个行业{金融业、采矿业、文化娱乐}，则金融业编码为{1，0，0}，采矿业编码为{0，1，0}，文化娱乐编码为{0，0，1}。既会增加特征向量，也就是获取到用户账号的当前特征值。并对当前特征值和各个历史特征值进行标准化特征，且方式均相同。And after the enterprise logs in based on the user account on the platform, that is, for the enterprise that logs in to the platform, the similarity calculation will be performed according to the current latest selection characteristics of the enterprise and the characteristics of each successfully applied enterprise. The calculation steps are as follows: For the data of the user account The information is encoded as discrete data, that is, for the ordered sequence {A->B->C->D} like taxpayer grades, it is converted into an ordered sequence such as {1, 2, 3, 4}; for the unordered sequence such as industry , using one-hot coding, that is, assuming that there are 3 industries {financial industry, mining industry, cultural entertainment}, the financial industry coding is {1, 0, 0}, the mining industry coding is {0, 1, 0}, and the cultural industry coding is {0, 1, 0}. Recreational codes are {0, 0, 1}. Not only will the feature vector be added, that is, the current feature value of the user account will be obtained. Standardize the current eigenvalue and each historical eigenvalue in the same way.
例如，对表2中每个特征采用z-score(标准)方式进行标准化，标准化后的值为
计算企业当前标准化特征值与其它企业申请成功时的历史标准化特征值的相似度S，相似度采用欧式距离进行计算，距离越小，相似度越高。Calculate the similarity S between the current standardized eigenvalues of the enterprise and the historical standardized eigenvalues of other enterprises when the application is successful. The similarity is calculated using the Euclidean distance. The smaller the distance, the higher the similarity.
其中n为标准化特征的个数；当计算出各个相似度后，选取与企业相似度最高的K(预设数量)个申请成功资源信息时的企业特征(即获取预设数量的目标相似度值)，其对应的企业相似度为S(u，k)，其对应的资源信息为L(L＝{L1，L2，L3……Ln})。Among them, n is the number of standardized features; after calculating each similarity, select the K (preset number) enterprise features with the highest similarity to the enterprise when applying for successful resource information (that is, to obtain a preset number of target similarity values) ), its corresponding enterprise similarity is S(u, k), and its corresponding resource information is L (L={L1, L2, L3...Ln}).
步骤S30，基于各所述资源信息确定各所述目标相似度值对应的目标资源信息，并获取各所述目标资源信息中的参数信息；Step S30, determining target resource information corresponding to each of the target similarity values based on each of the resource information, and acquiring parameter information in each of the target resource information;
当获取到各个目标相似度值后，还需要获取各个资源信息的参数信息，如当资源信息为金融资源信息时，则参数信息可以包括金融资源信息的申请额度和申请时间。并在各个资源信息中确定与各个目标相似度值对应的目标资源信息，也就是目标资源信息的数量和目标相似度值的数量是相同的，并通过网络或者历史数据等方式获取各个目标资源信息中的参数信息。其中，参数信息包括资源信息的额度限制，时间限制等。After each target similarity value is obtained, parameter information of each resource information needs to be obtained. For example, when the resource information is financial resource information, the parameter information may include the application amount and application time of the financial resource information. And determine the target resource information corresponding to each target similarity value in each resource information, that is, the number of target resource information and the number of target similarity values are the same, and obtain each target resource information through network or historical data, etc. parameter information in . Wherein, the parameter information includes a quota limit of resource information, a time limit, and the like.
步骤S40，基于预设推荐算法对各所述目标相似度值和各所述参数信息进行计算，以获取各所述资源信息对应的推荐指数，并在各所述推荐指数中确定数值最高的目标推荐指数，对所述目标推荐指数对应的资源信息进行推荐。Step S40: Calculate the similarity value of each of the targets and each of the parameter information based on a preset recommendation algorithm, so as to obtain a recommendation index corresponding to each of the resource information, and determine the target with the highest value in each of the recommended indices A recommendation index, for recommending resource information corresponding to the target recommendation index.
资源信息的推荐指数除了相似度指数外，由于企业申请到的额度越高，说明越符合资源信息的要求，因此和额度高的企业特征相似会更有机会申请成功，同时由于资源信息根据经济周期也在变化，因此时间越早，参考性也越弱，因此对企业U和上一步骤产生的L中每个资源信息的推荐指数C(U，Li)计算公式如下：In addition to the similarity index for the recommendation index of resource information, because the higher the quota applied by the enterprise, the more it meets the requirements of the resource information, so the characteristics of the enterprise with the high quota will have a better chance of successful application. At the same time, because the resource information is based on the economic cycle is also changing, so the earlier the time, the weaker the reference. Therefore, the calculation formula of the recommendation index C(U, Li) for each resource information in the enterprise U and L generated in the previous step is as follows:
其中m为K企业特征申请的信贷额度；
另外，为辅助理解本申请中的资源信息推荐方法的流程，下面进行举例说明。In addition, in order to assist in understanding the flow of the resource information recommendation method in the present application, an example is described below.
例如，如图4所示，目前企业先在金融平台汇总进行申请，并通过金融平台进行数据转发(转发数据包括企业的申请信息)到金融机构，金融机构根据金融平台提供的数据对企业进行线上线下提供金融服务。单这样操作导致企业申请的成功率很低。而在本实施例中，则可以提高企业申请的成功率，如图5所示，先进行特征选择，即根据平台企业基本信息及成功申请的历史数据，通过特征选择筛选出独立且相关性较大的特征。再进行相似度计算，即使用企业当前特征信息计算企业与各资源信息成功申请时的特征的相似度，并筛选出相似度最高的K个企业申请时特征及对应的L个资源信息；再进行推荐指数计算，即根据成功申请的额度及申请时间计算这L个资源信息的推荐指数，选取推荐指数最高的信贷进行推荐；并进行效果定期评估，即跟踪推荐效果，并对特征进行定期更新，再每隔一段时间进行重新特征选取。For example, as shown in Figure 4, at present, the enterprise first collects the application on the financial platform, and forwards the data through the financial platform (the forwarded data includes the application information of the enterprise) to the financial institution. Provide financial services online and offline. This operation alone results in a very low success rate for business applications. In this embodiment, the success rate of enterprise application can be improved. As shown in FIG. 5, feature selection is performed first, that is, according to the basic information of the platform enterprise and the historical data of successful applications, independent and more relevant ones are selected through feature selection. big feature. Then carry out the similarity calculation, that is, use the current feature information of the enterprise to calculate the similarity of the features when the enterprise and each resource information successfully applied, and screen out the K enterprise application features with the highest similarity and the corresponding L resource information; Recommendation index calculation, that is, calculate the recommendation index of the L resource information according to the amount of successful application and application time, and select the credit with the highest recommendation index for recommendation; and conduct regular evaluation of the effect, that is, track the recommendation effect and update the characteristics regularly Re-feature selection is performed at regular intervals.
在本实施例中，通过获取各资源信息的历史数据，并基于特征选择在各所述历史数据中获取与用户账号具有相关性的多个历史特征值；计算用户账号的当前特征值与各所述历史特征值之间的相似度值，并在各所述相似度值中获取预设数量的目标相似度值；基于各所述资源信息确定各所述目标相似度值对应的目标资源信息，并获取各所述目标资源信息中的参数信息；基于预设推荐算法对各所述目标相似度值和各所述参数信息进行计算，以获取各所述资源信息对应的推荐指数，并在各所述推荐指数中确定数值最高的目标推荐指数，对所述目标推荐指数对应的资源信息进行推荐。通过获取各个资源信息的历史数据，并将用户账号的当前特征值和历史数据中的多个历史特征值进行比较，以确定相似度值，也就是在计算相似度值时，综合考虑了资源信息的历史数据，提高了向用户账号推荐资源信息的准确性，并且还需要在这些相似度值中确定目标相似度值，并根据参数信息确定目标推荐指数，将目标推荐指数对应的资源信息进行推荐，使得本次推荐的效果更好，更符合用户的需求，提高了金融平台中推荐算法推荐的准确性。In this embodiment, the historical data of each resource information is obtained, and a plurality of historical feature values relevant to the user account are obtained in each of the historical data based on the feature selection; the current feature value of the user account is calculated and each The similarity value between the historical feature values is obtained, and a preset number of target similarity values are obtained in each of the similarity values; the target resource information corresponding to each of the target similarity values is determined based on each of the resource information, and obtain the parameter information in each of the target resource information; calculate each of the target similarity values and each of the parameter information based on a preset recommendation algorithm, to obtain the recommendation index corresponding to each of the resource information, and in each The target recommendation index with the highest value is determined in the recommendation indices, and the resource information corresponding to the target recommendation index is recommended. By acquiring the historical data of each resource information, and comparing the current feature value of the user account with multiple historical feature values in the historical data, the similarity value is determined, that is, the resource information is comprehensively considered when calculating the similarity value. It improves the accuracy of recommending resource information to user accounts, and also needs to determine the target similarity value among these similarity values, and determine the target recommendation index according to the parameter information, and recommend the resource information corresponding to the target recommendation index. , so that the effect of this recommendation is better, more in line with the needs of users, and improves the accuracy of the recommendation algorithm in the financial platform.
进一步地，基于本发明资源信息推荐方法第一实施例，提出本发明资源信息推荐方法第二实施例。本实施例是本发明第一实施例的步骤S40，基于预设推荐算法对各所述目标相似度值和各所述参数信息进行计算，以获取各所述资源信息对应的推荐指数的步骤的细化，包括：Further, based on the first embodiment of the resource information recommendation method of the present invention, a second embodiment of the resource information recommendation method of the present invention is proposed. This embodiment is step S40 of the first embodiment of the present invention, calculating each of the target similarity values and each of the parameter information based on a preset recommendation algorithm to obtain a recommendation index corresponding to each of the resource information. refinement, including:
步骤a，依次遍历各所述目标相似度值，并确定当前遍历的当前目标相似度值对应的参数信息；Step a, traverse each of the target similarity values in turn, and determine the parameter information corresponding to the current target similarity value currently traversed;
在获取到各个目标相似度值和各个参数信息后，依次遍历各个目标相似度值，并在各个参数信息中确定当前遍历的当前目标相似度值对应的参数信息。After each target similarity value and each parameter information are acquired, each target similarity value is traversed in turn, and parameter information corresponding to the currently traversed current target similarity value is determined in each parameter information.
步骤b，根据预设推荐算法对所述当前目标相似度值和所述当前目标相似度值对应的参数信息进行计算，以获取所述当前目标相似度值对应的资源信息的推荐指数，直至各所述目标相似度值遍历完成。Step b: Calculate the current target similarity value and the parameter information corresponding to the current target similarity value according to a preset recommendation algorithm, so as to obtain the recommendation index of the resource information corresponding to the current target similarity value. The target similarity value traversal is completed.
通过预设推荐指数计算公式来对当前目标相似度值、当前目标相似度值对应的参数信息(如申请额度和申请时间)进行计算，以获取到当前目标相似度值对应的资源信息的推荐指数，并对所有目标相似度采用同样的方式进行计算，得到各个目标相似度对应的资源信息的推荐指数。其中预设推荐指数计算公式可以是：Calculate the current target similarity value and the parameter information (such as application quota and application time) corresponding to the current target similarity value by using the preset recommendation index calculation formula, so as to obtain the recommendation index of the resource information corresponding to the current target similarity value , and calculate the similarity of all targets in the same way to obtain the recommendation index of the resource information corresponding to the similarity of each target. The preset recommendation index calculation formula can be:
其中m为K企业特征申请的信贷额度；为L个资源信息最大贷款额度的平均值；y为K企业特征申请时的距离10年前的时间长度，以年为单位，取2位小数；S(u，k)为上一步获得的企业与K企业特征的相似度，Li为上一步获得的资源信息集合L中的资源信息；N(L)为Li资源信息对应的K个企业特征个数；计算公式主要的含义是对资源信息集合L，可能多个企业特征Ki会对用同一个资源信息Li，按公式计算资源信息Li对应的多个企业特征Ki的推荐指数并汇总，就是用户对于资源信息Li的推荐指数，最后选择推荐指数最高的D(预设数量)个资源信息进行推荐。也就是对每个资源信息(即资源信息)均采用相同的方式进行计算得到各个推荐指数。Among them, m is the credit limit applied for by the characteristics of K enterprise; is the average value of the maximum loan amount of L resource information; y is the length of time from 10 years ago when K enterprise characteristics applied, in years, with 2 decimal places; S(u, k) is the enterprise obtained in the previous step Similarity with K enterprise characteristics, Li is the resource information in the resource information set L obtained in the previous step; N(L) is the number of K enterprise characteristics corresponding to Li resource information; the main meaning of the calculation formula is the resource information set. L, it is possible that multiple enterprise features Ki will use the same resource information Li, calculate the recommendation index of multiple enterprise features Ki corresponding to the resource information Li according to the formula and summarize, which is the user's recommendation index for the resource information Li, and finally select the recommendation index The highest D (preset number) pieces of resource information are recommended. That is, each recommendation index is obtained by calculating each resource information (ie, resource information) in the same manner.
在本实施例中，通过根据预设推荐指数计算公式、目标相似度值、参数信息计算推荐指数，从而保障了获取到的推荐指数的准确性和有效性。In this embodiment, the recommendation index is calculated according to the preset recommendation index calculation formula, the target similarity value, and the parameter information, thereby ensuring the accuracy and validity of the obtained recommendation index.
进一步地，计算用户账号的当前特征值与各所述历史特征值之间的相似度值的步骤，包括：Further, the step of calculating the similarity value between the current feature value of the user account and each of the historical feature values includes:
步骤d，获取所述当前特征值对应的当前标准化特征值和各所述历史特征值对应的历史标准化特征值；Step d, obtaining the current standardized feature value corresponding to the current feature value and the historical standardized feature value corresponding to each of the historical feature values;
当获取到当前特征值后，可以采用如z-score方式对当前当前特征值进行标准化，以得到当前特征值对应的当前标准化特征值，并采用同样的方式如z-score方式对各个历史特征值进行标准化，得到各个历史特征值对应的历史标准化特征值。After the current eigenvalues are obtained, the current eigenvalues can be standardized by means such as z-score to obtain the current normalized eigenvalues corresponding to the current eigenvalues, and the same method such as z-score can be used for each historical eigenvalue. Carry out standardization to obtain the historical standardized feature value corresponding to each historical feature value.
步骤e，根据所述当前标准化特征值和各所述历史标准化特征值确定所有相似度值。Step e: Determine all similarity values according to the current standardized feature value and each of the historical standardized feature values.
当获取到当前标准化特征值和各个历史标准化特征值后，可以根据欧式距离来计算当前标准化特征值和各个历史标准化特征值之间的相似度值。After the current standardized feature value and each historical standardized feature value are obtained, the similarity value between the current standardized feature value and each historical standardized feature value can be calculated according to the Euclidean distance.
在本实施例中，通过获取当前标准化特征值和各个历史标准化特征值，再获取到各个相似度值，从而保障了获取到的相似度值的准确性。In this embodiment, each similarity value is obtained by obtaining the current standardized feature value and each historical standardized feature value, thereby ensuring the accuracy of the obtained similarity value.
具体地，获取所述当前特征值对应的当前标准化特征值和各所述历史特征值对应的历史标准化特征值的步骤，包括：Specifically, the step of obtaining the current standardized feature value corresponding to the current feature value and the historical standardized feature value corresponding to each of the historical feature values includes:
步骤f，基于预设数据处理方式对所述当前特征值进行标准化，以获取所述当前特征值对应的当前标准化特征值；Step f, standardizing the current feature value based on a preset data processing method to obtain the current standardized feature value corresponding to the current feature value;
当获取到当前特征值后，采用预设数据处理方式对当前特征值进行标准化，以获取当前特征值对应的当前标准化特征值，如采用z-score方式进行。After the current eigenvalue is obtained, the current eigenvalue is standardized by a preset data processing method to obtain the current normalized eigenvalue corresponding to the current eigenvalue, for example, a z-score method is used.
步骤h，基于所述预设数据处理方式对各所述历史特征值进行标准化，以获取各所述历史特征值对应的历史标准化特征值。In step h, each of the historical feature values is standardized based on the preset data processing method to obtain a historical standardized feature value corresponding to each of the historical feature values.
需要说明的是，在对当前特征值进行标准化的方式和对各个历史特征值进行标准化的方式相同，也就是同样采用预设数据处理方式对各个历史特征值进行标准化，得到各个历史特征值对应的历史标准化特征值。It should be noted that the method of standardizing the current feature value is the same as the method of standardizing each historical feature value, that is, the preset data processing method is also used to standardize each historical feature value, and the corresponding historical feature value is obtained. Historical normalized eigenvalues.
在本实施例中，通过根据预设数据处理方式对当前特征值进行标准化，并对各个历史特征值进行标准化，从而保障了获取到的当前标准化特征值和历史标准化特征值的准确性。In this embodiment, by standardizing the current feature value according to the preset data processing method, and standardizing each historical feature value, the accuracy of the acquired current standardized feature value and historical standardized feature value is guaranteed.
具体地，根据所述当前标准化特征值和各所述历史标准化特征值确定所有相似度值的步骤，包括：Specifically, the step of determining all similarity values according to the current standardized feature value and each of the historical standardized feature values includes:
步骤k，依次遍历各所述历史标准化特征值，并计算当前遍历的历史标准化特征值和所述当前标准化特征值之间的欧式距离；Step k, traverse each of the historical standardized eigenvalues in turn, and calculate the Euclidean distance between the currently traversed historical standardized eigenvalues and the current standardized eigenvalues;
当获取到各个历史标准化特征值后，可以依次遍历各个历史标准化特征值，并计算当前遍历的历史标准化特征值和当前标准化特征值之间的欧式距离。After each historical standardized feature value is obtained, each historical standardized feature value can be traversed in turn, and the Euclidean distance between the currently traversed historical standardized feature value and the current standardized feature value is calculated.
步骤m，基于所述欧式距离确定当前遍历的历史标准化特征值和当前标准化特征值的相似度值，直至各所述历史标准化特征值遍历完成。Step m: Determine, based on the Euclidean distance, the currently traversed historical standardized feature value and the similarity value of the current standardized feature value, until the traversal of each of the historical standardized feature values is completed.
根据计算得到的欧式距离来确定当前遍历的历史标准化特征值和当前标准化特征值之间的相似度值，可以是欧式距离越小，两者之间的相似度值就越高。直至获取到所有的历史标准化特征值和当前标准化特征值的相似度值。并且对所有历史标准化特征值和打过去标准化特征值之间的相似度值计算方式相同。The similarity value between the currently traversed historical standardized feature value and the current standardized feature value is determined according to the calculated Euclidean distance. The smaller the Euclidean distance, the higher the similarity value between the two. Until the similarity values of all historical standardized feature values and current standardized feature values are obtained. And the similarity value between all historical normalized eigenvalues and past normalized eigenvalues is calculated in the same way.
在本实施例中，通过计算当前标准化特征值和历史标准化特征值之间的欧式距离，并根据欧式距离计算相似度值，从而保障了计算得到的相似度值的准确性。In this embodiment, by calculating the Euclidean distance between the current standardized feature value and the historical standardized feature value, and calculating the similarity value according to the Euclidean distance, the accuracy of the calculated similarity value is guaranteed.
进一步地，基于本发明资源信息推荐方法第一至第二任意一个实施例的基础上，提出本发明资源信息推荐方法第三实施例。本实施例是本发明第一实施例的步骤S10，基于特征选择在各所述历史数据中获取与用户账号具有相关性的多个历史特征值的步骤的细化，包括：Further, based on any one of the first to second embodiments of the resource information recommendation method of the present invention, a third embodiment of the resource information recommendation method of the present invention is proposed. This embodiment is step S10 of the first embodiment of the present invention, the refinement of the step of acquiring a plurality of historical feature values related to the user account in each of the historical data based on feature selection, including:
步骤n，基于预设校验算法对各所述历史数据进行离散化，以获取各所述历史数据对应的初级特征值；Step n, discretizing each of the historical data based on a preset verification algorithm to obtain the primary characteristic value corresponding to each of the historical data;
当获取到各个历史数据后，可以通过预设校验算法(如卡方校验方法)对这些历史数据进行离散化，如对上年营业收入字段离散化后为：小于等于3000万为1，大于3000万并且小于等于1亿为2，大于1亿并且小于等于10亿为4，当对这些历史数据进行离散化后得到的各个结果就是各个历史数据对应的初级特征值。After each historical data is obtained, these historical data can be discretized by a preset verification algorithm (such as the chi-square verification method). Greater than 30 million and less than or equal to 100 million is 2, greater than 100 million and less than or equal to 1 billion is 4, and each result obtained after discretizing these historical data is the primary eigenvalue corresponding to each historical data.
其中，卡方检验是一种假设性检验的方法，它能够检验两个分类变量之间是否是独立无关的。它通过观察实际值和理论值的偏差来确定原假设是否成立。Among them, the chi-square test is a hypothesis testing method, which can test whether two categorical variables are independent and unrelated. It determines whether the null hypothesis holds by observing the deviation of the actual value from the theoretical value.
步骤x，基于所述预设校验算法对各所述初级特征值进行评分，并基于评分结果获取与用户账号具有相关性的多个历史特征值。In step x, each of the primary feature values is scored based on the preset verification algorithm, and a plurality of historical feature values related to the user account are acquired based on the scoring results.
当获取到各个初级特征值后，还可以采用过滤法中的卡方校验法(即预设校验算法)对每个初级特征值进行评分，并根据评分结果选取一定数量的评分较高的初级特征值作为与用户具有相关性的多个历史特征值。After each primary eigenvalue is obtained, the chi-square verification method in the filtering method (ie, the preset verification algorithm) can also be used to score each primary eigenvalue, and a certain number of high-scoring ones are selected according to the scoring results. The primary eigenvalues serve as multiple historical eigenvalues that are relevant to the user.
在本实施例中，通过根据预设校验算法对各个历史数据进行离散化得到各个初级特征值，并对这些初级特征值进行评分，基于评分结果获取历史特征值，从而保障了获取到的历史特征值的准确性。In this embodiment, each primary eigenvalue is obtained by discretizing each historical data according to a preset verification algorithm, and these primary eigenvalues are scored, and the historical eigenvalue is obtained based on the scoring result, thereby ensuring the acquired history Accuracy of eigenvalues.
进一步地，计算用户账号的当前特征值与各所述历史特征值之间的相似度值的步骤之前，包括：Further, before the step of calculating the similarity value between the current feature value of the user account and each of the historical feature values, the steps include:
步骤y，获取用户账号的当前数据，并对所述当前数据进行离散数据编码，以获取所述当前账号的当前特征值。In step y, the current data of the user account is obtained, and discrete data encoding is performed on the current data to obtain the current feature value of the current account.
当企业通过用户账号登录平台后，会自动获取用户账号的当前数据，并对这些当前数据进行离散数据编码，以得到当前账号的当前特征值。如对于像纳税人等级的有序序列{A->B->C->D}转换为有序数列如{1、2、3、4}；对于行业等无序序列，采用one-hot编码，即假设有3个行业{金融业、采矿业、文化娱乐}，则金融业编码为{1，0，0}，采矿业编码为{0，1，0}，文化娱乐编码为{0，0，1}。既会增加特征向量，得到当前账号的当前特征值。When the enterprise logs in to the platform through the user account, it will automatically obtain the current data of the user account, and encode the current data discretely to obtain the current feature value of the current account. For example, for the ordered sequence {A->B->C->D} like taxpayer grades, convert it to an ordered sequence such as {1, 2, 3, 4}; for unordered sequences such as industries, use one-hot encoding , that is, assuming there are 3 industries {financial industry, mining industry, cultural and entertainment industry}, the financial industry code is {1, 0, 0}, the mining industry code is {0, 1, 0}, the cultural and entertainment industry code is {0, 0, 1}. It will increase the eigenvector and get the current eigenvalue of the current account.
在本实施例中，通过对当前数据进行离散数据编码，得到当前账号的当前特征值，从而保障了获取到的当前特征值的准确性。In this embodiment, the current feature value of the current account is obtained by performing discrete data encoding on the current data, thereby ensuring the accuracy of the obtained current feature value.
本发明实施例还提供一种资源信息推荐装置，参照图3，所述资源信息推荐装置包括：An embodiment of the present invention further provides an apparatus for recommending resource information. Referring to FIG. 3 , the apparatus for recommending resource information includes:
获取模块，用于获取各资源信息的历史数据，并基于特征选择在各所述历史数据中获取与用户账号具有相关性的多个历史特征值；an acquisition module, configured to acquire historical data of each resource information, and to acquire a plurality of historical feature values relevant to the user account in each of the historical data based on feature selection;
计算模块，用于计算用户账号的当前特征值与各所述历史特征值之间的相似度值，并在各所述相似度值中获取预设数量的目标相似度值；a calculation module, configured to calculate the similarity value between the current feature value of the user account and each of the historical feature values, and obtain a preset number of target similarity values from each of the similarity values;
确定模块，用于基于各所述资源信息确定各所述目标相似度值对应的目标资源信息，并获取各所述目标资源信息中的参数信息；a determining module, configured to determine target resource information corresponding to each of the target similarity values based on each of the resource information, and obtain parameter information in each of the target resource information;
推荐模块，用于基于预设推荐算法对各所述目标相似度值和各所述参数信息进行计算，以获取各所述资源信息对应的推荐指数，并在各所述推荐指数中确定数值最高的目标推荐指数，对所述目标推荐指数对应的资源信息进行推荐。A recommendation module, configured to calculate each of the target similarity values and each of the parameter information based on a preset recommendation algorithm, to obtain a recommendation index corresponding to each of the resource information, and to determine the highest value in each of the recommended indices The target recommendation index is the target recommendation index, and the resource information corresponding to the target recommendation index is recommended.
可选地，所述推荐模块，还用于：Optionally, the recommendation module is further used for:
依次遍历各所述目标相似度值，并确定当前遍历的当前目标相似度值对应的参数信息；Traversing each of the target similarity values in turn, and determining the parameter information corresponding to the current target similarity value currently traversed;
根据预设推荐算法对所述当前目标相似度值和所述当前目标相似度值对应的参数信息进行计算，以获取所述当前目标相似度值对应的资源信息的推荐指数，直至各所述目标相似度值遍历完成。Calculate the current target similarity value and the parameter information corresponding to the current target similarity value according to the preset recommendation algorithm, so as to obtain the recommendation index of the resource information corresponding to the current target similarity value, until each target The similarity value traversal is completed.
可选地，所述计算模块，还用于：Optionally, the computing module is also used for:
获取所述当前特征值对应的当前标准化特征值和各所述历史特征值对应的历史标准化特征值；Obtain the current standardized feature value corresponding to the current feature value and the historical standardized feature value corresponding to each of the historical feature values;
根据所述当前标准化特征值和各所述历史标准化特征值确定所有相似度值。All similarity values are determined according to the current normalized feature value and each of the historical normalized feature values.
可选地，所述计算模块，还用于：Optionally, the computing module is also used for:
基于预设数据处理方式对所述当前特征值进行标准化，以获取所述当前特征值对应的当前标准化特征值；Standardize the current feature value based on a preset data processing method to obtain a current standardized feature value corresponding to the current feature value;
基于所述预设数据处理方式对各所述历史特征值进行标准化，以获取各所述历史特征值对应的历史标准化特征值。Each of the historical feature values is normalized based on the preset data processing method to obtain a historical standardized feature value corresponding to each of the historical feature values.
可选地，所述计算模块，还用于：Optionally, the computing module is also used for:
依次遍历各所述历史标准化特征值，并计算当前遍历的历史标准化特征值和所述当前标准化特征值之间的欧式距离；Traversing each of the historical standardized eigenvalues in turn, and calculating the Euclidean distance between the currently traversed historical standardized eigenvalues and the current standardized eigenvalues;
基于所述欧式距离确定当前遍历的历史标准化特征值和当前标准化特征值的相似度值，直至各所述历史标准化特征值遍历完成。Based on the Euclidean distance, the currently traversed historical normalized feature value and the similarity value of the current normalized feature value are determined until each historical normalized feature value traversal is completed.
可选地，所述获取模块，还用于：Optionally, the obtaining module is also used for:
基于预设校验算法对各所述历史数据进行离散化，以获取各所述历史数据对应的初级特征值；Discretize each of the historical data based on a preset verification algorithm to obtain the primary characteristic value corresponding to each of the historical data;
基于所述预设校验算法对各所述初级特征值进行评分，并基于评分结果获取与用户账号具有相关性的多个历史特征值。Each of the primary feature values is scored based on the preset verification algorithm, and a plurality of historical feature values related to the user account are acquired based on the scoring result.
所述资源信息推荐装置，还包括：The resource information recommendation device further includes:
获取用户账号的当前数据，并对所述当前数据进行离散数据编码，以获取所述当前账号的当前特征值。The current data of the user account is obtained, and discrete data encoding is performed on the current data to obtain the current feature value of the current account.
上述各程序模块所执行的方法可参照本发明资源信息推荐方法各个实施例，此处不再赘述。For the methods executed by the above program modules, reference may be made to the various embodiments of the resource information recommendation method of the present invention, which will not be repeated here.
本发明还提供一种计算机存储介质。The present invention also provides a computer storage medium.
本发明计算机存储介质上存储有资源信息推荐程序，所述资源信息推荐程序被处理器执行时实现如上所述的资源信息推荐方法的步骤。A resource information recommendation program is stored on the computer storage medium of the present invention, and when the resource information recommendation program is executed by the processor, the steps of the resource information recommendation method described above are implemented.
其中，在所述处理器上运行的资源信息推荐程序被执行时所实现的方法可参照本发明资源信息推荐方法各个实施例，此处不再赘述。For the method implemented when the resource information recommendation program running on the processor is executed, reference may be made to the various embodiments of the resource information recommendation method of the present invention, which will not be repeated here.
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者系统不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者系统所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、方法、物品或者系统中还存在另外的相同要素。It should be noted that, herein, the terms "comprising", "comprising" or any other variation thereof are intended to encompass non-exclusive inclusion, such that a process, method, article or system comprising a series of elements includes not only those elements, It also includes other elements not expressly listed or inherent to such a process, method, article or system. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article or system that includes the element.
上述本发明实施例序号仅仅为了描述，不代表实施例的优劣。The above-mentioned serial numbers of the embodiments of the present invention are only for description, and do not represent the advantages or disadvantages of the embodiments.
通过以上的实施方式的描述，本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件，但很多情况下前者是更佳的实施方式。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品存储在如上所述的一个存储介质(如ROM/RAM、磁碟、光盘)中，包括若干指令用以使得一台终端设备(可以是手机，计算机，服务器，空调器，或者网络设备等)执行本发明各个实施例所述的方法。From the description of the above embodiments, those skilled in the art can clearly understand that the method of the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is better implementation. Based on such understanding, the technical solutions of the present invention can be embodied in the form of software products in essence or the parts that make contributions to the prior art, and the computer software products are stored in a storage medium (such as ROM/RAM) as described above. , magnetic disk, optical disk), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the methods described in the various embodiments of the present invention.
以上仅为本发明的优选实施例，并非因此限制本发明的专利范围，凡是利用本发明说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本发明的专利保护范围内。The above are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied in other related technical fields , are similarly included in the scope of patent protection of the present invention.
Claims (10)
Translated fromChinesePriority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911145222.1ACN110851729A (en) | 2019-11-19 | 2019-11-19 | Resource information recommendation method, apparatus, device and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911145222.1ACN110851729A (en) | 2019-11-19 | 2019-11-19 | Resource information recommendation method, apparatus, device and computer storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110851729Atrue CN110851729A (en) | 2020-02-28 |
Family
ID=69603483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911145222.1APendingCN110851729A (en) | 2019-11-19 | 2019-11-19 | Resource information recommendation method, apparatus, device and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110851729A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111984864A (en)* | 2020-08-14 | 2020-11-24 | 北京达佳互联信息技术有限公司 | Object recommendation method and device, electronic equipment and storage medium |
| CN112016975A (en)* | 2020-09-02 | 2020-12-01 | 中国平安财产保险股份有限公司 | Product screening method and device, computer equipment and readable storage medium |
| CN112100489A (en)* | 2020-08-27 | 2020-12-18 | 北京百度网讯科技有限公司 | Method, apparatus and computer storage medium for subject recommendation |
| CN112182052A (en)* | 2020-10-12 | 2021-01-05 | 中国铁塔股份有限公司 | A parameter display method, apparatus, electronic device and readable storage medium |
| CN112464106A (en)* | 2020-11-26 | 2021-03-09 | 上海哔哩哔哩科技有限公司 | Object recommendation method and device |
| CN112613426A (en)* | 2020-12-26 | 2021-04-06 | 北京国腾联信科技有限公司 | Resource area determination method, device, equipment and storage medium |
| CN113836391A (en)* | 2020-06-24 | 2021-12-24 | 北京达佳互联信息技术有限公司 | Resource recommendation method and device, computer equipment and storage medium |
| CN114119069A (en)* | 2021-10-29 | 2022-03-01 | 大箴(杭州)科技有限公司 | Resource management product recommendation method and device, electronic equipment and storage medium |
| CN116227826A (en)* | 2022-12-23 | 2023-06-06 | 车主邦（北京）科技有限公司 | Resource distribution method, device, storage medium and computer equipment |
| CN116955736A (en)* | 2023-09-15 | 2023-10-27 | 北京南天智联信息科技股份有限公司 | Data constraint condition recommendation method and system in data standard |
| CN119255021A (en)* | 2024-12-06 | 2025-01-03 | 杭州海康威视系统技术有限公司 | Point resource recommendation method, device, and electronic device |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992181A (en)* | 2015-06-29 | 2015-10-21 | 昆明理工大学 | Method for recommending books in real time according to habits of library user |
| CN107220852A (en)* | 2017-05-26 | 2017-09-29 | 北京小度信息科技有限公司 | Method, device and server for determining target recommended user |
| CN108573032A (en)* | 2018-03-27 | 2018-09-25 | 麒麟合盛网络技术股份有限公司 | Video recommendation method and device |
| CN109214856A (en)* | 2018-07-27 | 2019-01-15 | 百度在线网络技术（北京）有限公司 | The method for digging and device, computer equipment and readable medium that user is intended to |
| CN109241415A (en)* | 2018-08-20 | 2019-01-18 | 平安科技（深圳）有限公司 | Item recommendation method, device, computer equipment and storage medium |
| CN109426985A (en)* | 2017-09-04 | 2019-03-05 | 腾讯科技（深圳）有限公司 | Information processing method, device, computer equipment and storage medium |
| CN109508407A (en)* | 2019-01-14 | 2019-03-22 | 上海电机学院 | The tv product recommended method of time of fusion and Interest Similarity |
| CN110163476A (en)* | 2019-04-15 | 2019-08-23 | 重庆金融资产交易所有限责任公司 | Project intelligent recommendation method, electronic device and storage medium |
- 2019
- 2019-11-19CNCN201911145222.1Apatent/CN110851729A/enactivePending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104992181A (en)* | 2015-06-29 | 2015-10-21 | 昆明理工大学 | Method for recommending books in real time according to habits of library user |
| CN107220852A (en)* | 2017-05-26 | 2017-09-29 | 北京小度信息科技有限公司 | Method, device and server for determining target recommended user |
| CN109426985A (en)* | 2017-09-04 | 2019-03-05 | 腾讯科技（深圳）有限公司 | Information processing method, device, computer equipment and storage medium |
| CN108573032A (en)* | 2018-03-27 | 2018-09-25 | 麒麟合盛网络技术股份有限公司 | Video recommendation method and device |
| CN109214856A (en)* | 2018-07-27 | 2019-01-15 | 百度在线网络技术（北京）有限公司 | The method for digging and device, computer equipment and readable medium that user is intended to |
| CN109241415A (en)* | 2018-08-20 | 2019-01-18 | 平安科技（深圳）有限公司 | Item recommendation method, device, computer equipment and storage medium |
| CN109508407A (en)* | 2019-01-14 | 2019-03-22 | 上海电机学院 | The tv product recommended method of time of fusion and Interest Similarity |
| CN110163476A (en)* | 2019-04-15 | 2019-08-23 | 重庆金融资产交易所有限责任公司 | Project intelligent recommendation method, electronic device and storage medium |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113836391A (en)* | 2020-06-24 | 2021-12-24 | 北京达佳互联信息技术有限公司 | Resource recommendation method and device, computer equipment and storage medium |
| CN113836391B (en)* | 2020-06-24 | 2023-07-11 | 北京达佳互联信息技术有限公司 | Resource recommendation method, device, computer equipment and storage medium |
| CN111984864A (en)* | 2020-08-14 | 2020-11-24 | 北京达佳互联信息技术有限公司 | Object recommendation method and device, electronic equipment and storage medium |
| CN111984864B (en)* | 2020-08-14 | 2023-12-26 | 北京达佳互联信息技术有限公司 | Object recommendation method, device, electronic equipment and storage medium |
| CN112100489B (en)* | 2020-08-27 | 2022-07-15 | 北京百度网讯科技有限公司 | Method, apparatus and computer storage medium for subject recommendation |
| CN112100489A (en)* | 2020-08-27 | 2020-12-18 | 北京百度网讯科技有限公司 | Method, apparatus and computer storage medium for subject recommendation |
| CN112016975A (en)* | 2020-09-02 | 2020-12-01 | 中国平安财产保险股份有限公司 | Product screening method and device, computer equipment and readable storage medium |
| CN112182052A (en)* | 2020-10-12 | 2021-01-05 | 中国铁塔股份有限公司 | A parameter display method, apparatus, electronic device and readable storage medium |
| CN112464106A (en)* | 2020-11-26 | 2021-03-09 | 上海哔哩哔哩科技有限公司 | Object recommendation method and device |
| CN112464106B (en)* | 2020-11-26 | 2022-12-13 | 上海哔哩哔哩科技有限公司 | Object recommendation method and device |
| CN112613426A (en)* | 2020-12-26 | 2021-04-06 | 北京国腾联信科技有限公司 | Resource area determination method, device, equipment and storage medium |
| CN114119069A (en)* | 2021-10-29 | 2022-03-01 | 大箴(杭州)科技有限公司 | Resource management product recommendation method and device, electronic equipment and storage medium |
| CN116227826A (en)* | 2022-12-23 | 2023-06-06 | 车主邦（北京）科技有限公司 | Resource distribution method, device, storage medium and computer equipment |
| CN116955736A (en)* | 2023-09-15 | 2023-10-27 | 北京南天智联信息科技股份有限公司 | Data constraint condition recommendation method and system in data standard |
| CN116955736B (en)* | 2023-09-15 | 2023-12-01 | 北京南天智联信息科技股份有限公司 | Data constraint condition recommendation method and system in data standard |
| CN119255021A (en)* | 2024-12-06 | 2025-01-03 | 杭州海康威视系统技术有限公司 | Point resource recommendation method, device, and electronic device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110851729A (en) | Resource information recommendation method, apparatus, device and computer storage medium | |
| CN110119413B (en) | Data fusion method and device | |
| US11775412B2 (en) | Machine learning models applied to interaction data for facilitating modifications to online environments | |
| Azzeh | A replicated assessment and comparison of adaptation techniques for analogy-based effort estimation | |
| CN109711955B (en) | Poor evaluation early warning method and system based on current order and blacklist base establishment method | |
| US20230093756A1 (en) | Systems and methods for generating recommendations | |
| US10721266B1 (en) | Automated security incident remediation recommender | |
| US12051023B2 (en) | Benchmarking of user experience quality | |
| US20210357699A1 (en) | Data quality assessment for data analytics | |
| CN111754044A (en) | Employee behavior audit method, device, device and readable storage medium | |
| US12063325B1 (en) | Systems and methods for contextual communication between devices | |
| Zhang et al. | Precision Marketing Method of E‐Commerce Platform Based on Clustering Algorithm | |
| US20220058658A1 (en) | Method of scoring and valuing data for exchange | |
| CN116664306A (en) | Intelligent recommendation method and device for wind control rules, electronic equipment and medium | |
| CN112581281A (en) | Product recommendation method and device, storage medium and electronic equipment | |
| CN111210321B (en) | Risk early warning method and system based on contract management | |
| CN107424026A (en) | Businessman's reputation evaluation method and device | |
| US20220164374A1 (en) | Method of scoring and valuing data for exchange | |
| US20240152992A1 (en) | Data intelligence platform | |
| CN118504752A (en) | Determination method of transaction risk prediction model, transaction risk prediction method, device, equipment, storage medium and program product | |
| US12052230B2 (en) | Machine learning techniques for object authentication | |
| CN112906723A (en) | Feature selection method and device | |
| CN116304851A (en) | Data standard determining method, apparatus, device, medium and computer program product | |
| US10007398B2 (en) | Integrated supplier information tool | |
| CN113393267A (en) | Method, device and equipment for determining express logistics center position and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | Application publication date:20200228 | |
| RJ01 | Rejection of invention patent application after publication |